One thing some of the other answers have alluded to but not laid out explicitly is that even without legal interventions, without moral implications, and without the scammers explicitly barring your behavior, your strategy has a negative expected payout, at least it often does.
TLDR: If an optimal investor wouldn't make at least 172% in profits before the scam crashed, you won't be able to scam the scammer with your strategy.
Accepting some discretization errors, there's a surprisingly simple formula for the break-even interest rate in schemes like this (schemes offering periodic fixed interest which will pay out at any given point in time before a cliff, after which all money invested is lost).
If the scheme runs for n periods (e.g. n months) before exiting/bankrupting, and if your strategy is to invest for a single pay period, then for you to break even they need to offer an interest rate of 1/(n-1). If you prefer percentages, use 100/(n-1) instead.
For some sample payouts, consider what happens if the scammer offers 5% monthly and quits after a year. Your break even point was around 9%, so if you invested $1000 each month, you could expect to lose around $40 on average, or $480 over the course of the year. If instead you invested in the mining operation promising 20% monthly, you would profit a handsome $1320.
Improving the Strategy
Interestingly if the option is available to you, you're better off spending as little time with scammers as possible and making up for that in volume. Intuitively, the only "risk" in that model is the risk of rolling over their bankruptcy point, and by splitting a transaction that would roll over into several that won't and one that will you're increasing your overall profits.
That said, there's a fundamental limit at which no amount of splitting into smaller transactions will help. Imagine a hypothetical "investor" at day 0, and consider their earnings over the entire course of the scam. The break-even point for having a chance of scamming the scammers is when this ground-zero investor would multiply their profits by e~2.718. In other words, if the scam doesn't have high enough interest and a long enough runtime to generate 172% in profits for an optimal "investor", your strategy will lose money no matter how you play the game.
In other words, a Ponzi scheme can structure their scheme so that all attempts to scam them back with your technique will fail, even if every participant is trying to scam the scammer (standard assumptions about reasonably even distributions through time and lack of collusion between counter-scammers).
For example, a perfect investor in the monthly trading venture you linked would make almost 80% just before the crash if it ran for a year, and while large and compelling it is a far cry from 172%. It would have to run for nearly 21 months (do you trust a scammer to stay in business that long? I think Madoff was an exception) for you to be able to scam them effectively. The mining venture on the other hand offers a maximum of 1.2^12-1=792% over the course of a year, and it offers huge potential for exploitation.
You mentioned alternating between scammers with a fixed $1000 each time. What's interesting is that there isn't fundamentally any difference between investing $1000 with the same investor twice and investing $1000 separately with two different scammers (at least, discounting the increasing likelihood of the scam going out of business as time goes on).
The perceived security from investing in multiple ventures with small amounts of capital is that you're less likely to go bankrupt, but a similar effect is at play when you split your bankroll into multiple small payments and continue to counter-scam the same individual in one-month increments. Even though you're virtually guaranteed to eventually lose an investment in the end, you're also guaranteed to make up for it and then some in the interim -- at least, if it's a winnable investment.
Going further, if the scammer is offering a winning proposition (>172% profits) then your investments are more stable in the sense of having lower variance if you continue to invest the same small amounts in a single venture until they go bankrupt. If you randomly choose scammers and points in time to invest in them, even if all the investments could have been profitable (meeting that 172% threshold), you have the potential to get extremely unlucky and lose every bet at once. That can't happen if you transact with the same individual over and over, simply using small enough transactions to make the loss at the end manageable.
That isn't to say that you don't also want to diversify between scammers, but the reason to do so isn't to lessen the chance of losing everything in one go (getting hit with the blunt side of the inevitable fiscal cliff). The reason is to mitigate your losses from the chance that not all scammers will be in business long enough for you to hit that 172% threshold (keeping with our policy of blatantly ignoring legal effects, the potential for the scammer to retaliate and/or compensate, and so on).